Re: top-down balance purpose discussion -- resend

2014-01-23 Thread Alex Shi
Any more comments for this idea? :) On 01/22/2014 03:40 PM, Alex Shi wrote: > On 01/21/2014 10:57 PM, Peter Zijlstra wrote: >> On Tue, Jan 21, 2014 at 10:04:26PM +0800, Alex Shi wrote: >>> >>> Current scheduler load balance is bottom-up mode, each CPU need >>> initiate the balance by self. >>> >>

Re: top-down balance purpose discussion -- resend

2014-01-21 Thread Alex Shi
On 01/21/2014 10:57 PM, Peter Zijlstra wrote: > On Tue, Jan 21, 2014 at 10:04:26PM +0800, Alex Shi wrote: >> >> Current scheduler load balance is bottom-up mode, each CPU need >> initiate the balance by self. >> >> 1, Like in a integrate computer system, it has smt/core/cpu/numa, 4 >> level schedul

Re: top-down balance purpose discussion -- resend

2014-01-21 Thread Peter Zijlstra
On Tue, Jan 21, 2014 at 10:04:26PM +0800, Alex Shi wrote: > > Current scheduler load balance is bottom-up mode, each CPU need > initiate the balance by self. > > 1, Like in a integrate computer system, it has smt/core/cpu/numa, 4 > level scheduler domains. If there is just 2 tasks in whole system

top-down balance purpose discussion -- resend

2014-01-21 Thread Alex Shi
Current scheduler load balance is bottom-up mode, each CPU need initiate the balance by self. 1, Like in a integrate computer system, it has smt/core/cpu/numa, 4 level scheduler domains. If there is just 2 tasks in whole system that both running on cpu0. Current load balance need to pull task to